#Camryn Blawas BIOS 611 Final Project

##Introduction

My project will use climatology data from three NOAA buoys near Cape Hatteras, Duck, and Beaufort North Carolina to understand the trends in water temperature, specifically in relation to gulf stream changes. There was a heatwave in North Carolina in 2017 that may have been amplified by a Gulf Stream warm core ring.

To test this, we can compare water temperature above the Gulf Stream (Duck), at Gulf Stream shift (Cape Hatteras), and below the Gulf Stream (Beaufort). In theory, if the Gulf stream did release a warm core ring, there would be a rapid increase in water temperature most likely starting at the Duck buoy in early April, and confirmed intrusion of the later arrival of a surge of near bottom Cold Pool water (Gawarkiewicz et al., 2019). We can also compare rates of warming between these three locations and see the overall trends in water temperature on the NC Coast.

These datasets have a variety of climate based variables including air temperature, water temperature, wind speed, atmospheric pressure, and a few other variables that may be of interest. For the purposes of this project, we are particularly interested in water temperature, as it would be most indicative of an intrusion of a warm core ring. The stipulation to the hypothesis that the presence of a warm core ring would cause higher or lower water temperature is that warm core rings may cross the shelf, but seeing their presence in shallow coastal waters (like where these buoys are located) may be difficult due to the warming at the air-sea interface.

Measurements are taken every 6 minutes, and the dataset spans from 2016 to present day. To make these datasets easily usable, we will average measurements each day. There is a large gap in data at the Duck buoy due to water temperature sensors failing from December 2019 to October of 2021, which does not impact the results of the warm core ring hypothesis, but does affect the overall trend analysis.

Figure 1. Site Map

##Overall Trend Analysis

To understand any overall water temperature trends over time, I plotted linear models at each site (Figs. 2, 3, 4). From these models, there does seem to be a very slightly positive linear trend over time at each site. The most interesting trend from these models is that it seems that Beaufort’s water temperature may be rising faster than at the other sites. This is particularly important in a physical oceanography context, since recent research into destabilization of the Gulf Stream could be the culprit for the intrusion of warmer waters (Andres et al., 2016). Other factors that could cause an increase in Beaufort’s water temperature could be atmospheric warming, changes in inlet dynamics due to storm activity, and other stressors related to global climate change. To investigate further if there are any seasonal trends, I attempted to detrend the seasonal curve of the water temperature trend at each site. I used the “trend” and “tseries” package to take out the seasonal curve and perform a Mann-Kendall trend test to determine if there was a seasonal trend. If the p-values were below 0.05, they indicate that there is a seasonal trend (this should be true at all sites, as temperature is known to vary with the seasons). This proved to be quite difficult at Duck, but from the other sites we can begin to see some conclusions. At Duck, we can see that the functions had a difficult time teasing out trends with such a large data gap, so the linear fit pretty much mimics the seasonal trend (Figure 5). At Cape Hatteras, we can see that the detrended linear model shows slight variations (I’ve chosen to ignore the very first curve along with 2016, as I think it is not detrending well at the very beginning) after 2018, but nothing incredibly significant, and the p value indicates that the seasonal trend is present (expected) (Figure 6). At Beaufort, we see quite similar trends with small fluctuations, but nothing incredibly indicative of a warm core ring or major increases in temperature, and the p value indicates that the seasonal trend is present (Figure 7). From this, we can see that the trend in water temperature and in fact, seasonal, however there are not extreme fluctuations and this analysis needs to be refined. To use some dimensionality reduction skills I learned in class, I attempted a PCA and K Means clustering analysis of my data. From the PCA of all data at each site combined, I found that wind direction was contributing 99% of the variance to PC1, which is not exactly relevant to my overall analysis, but is interesting (Figure 8). I then performed a clustering, but found the results not indicative of much (Figure 9). The two clusters seem to be overlapping a lot, indicating I need to refine my method or that a clustering analysis is not appropriate for this data. My hope was that a clustering would result in points that were indicative of storm activity or a warm core ring, but that does not seem to be the case. I think with a different dataset I might be able to do this, which I plan on pursuing next semester.

Figure 2. Duck Sea Surface Temperature Linear Model

Figure 3. Cape Hatteras Sea Surface Temperature Linear Model

Figure 4. Beaufort Sea Surface Temperature Linear Model

Figure 5. Duck Seasonal Temperature Trend

Figure 6. Cape Hatteras Seasonal Temperature Trend

Figure 7. Beaufort Seasonal Temperature Trend

Figure 8. Biplot of All Data PCA

Figure 9. Clustering Plot of All Data

Warm Core Ring Analysis

To understand if there were differences in temperature during late April that would indicate the presence of a warm core ring, I decided to compare water temperatures in the month of April during 2017, as well as weeks where there could have been warm water from the warm core ring or cold water from the slope. I began by comparing mean water temperature between years at each site using Wilcoxon nonparametric tests, with adjusted p values. At Duck, there was a clear decrease in temperature from 2017 to 2018, but there was no difference between 2016 and 2017, nor 2019 and 2017, indicating that April was not significantly warmer at Duck (Duck Blue Boxplot). At Cape Hatteras, there were statistically significant differences between 2017 and 2016, 2018, 2020, and 2021, but not 2017 and 2019 (CH Blue Boxplot). This is interesting, because statistically higher temperatures could indicate intrusion of warm core ring waters, but a smaller scale is necessary to absolutely confirm this hypothesis. At Beaufort, there were statistically significant differences between 2017 and 2016 and 2018, but not 2017 and 2020 nor 2021 (Beaufort Blue Boxplot). This is less indicative of a significance of April 2017 in terms of increased water temperature because Beaufort is below the biogeographical barrier, so it is unlikely a warm core ring would proliferate that far (Hoarfrost et al., 2019). To get at a hopefully smaller scale, I binned the days of each week in the warm core ring’s possible path down the coast and compared their water temperatures. Using a Wilcoxon non parametric test, I found no difference between weeks of possible warm core ring presence (All WCR Boxplot). This demonstrates that it is not likely that there was a warm core ring that occurred specifically at these sites during this period to bring in a water mass of a vastly different temperature.

Figure 10. Duck April Trend Boxplot

Figure 11. Cape Hatteras April Trend Boxplot

Figure 12. Beaufort April Trend Boxplot

Figure 13. All WCR Weeks Boxplot

##References:

Andres, M. “On the recent destabilization of the Gulf Stream path downstream of Cape Hatteras.” Geophysical Research Letters 43.18 (2016): 9836-9842.

Hoarfrost A, Balmonte JP, Ghobrial S, Ziervogel K, Bane J, Gawarkiewicz G and Arnosti C (2019) Gulf Stream Ring Water Intrusion on the Mid-Atlantic Bight Continental Shelf Break Affects Microbially Driven Carbon Cycling. Front. Mar. Sci. 6:394. doi: 10.3389/fmars.2019.00394

Gawarkiewicz G, Chen K, Forsyth J, Bahr F, Mercer AM, Ellertson A, Fratantoni P, Seim H, Haines S and Han L (2019) Characteristics of an Advective Marine Heatwave in the Middle Atlantic Bight in Early 2017. Front. Mar. Sci. 6:712. doi: 10.3389/fmars.2019.00712